Comprehensive evaluation of data-driven mode decomposition techniques for the analysis of separating and reattaching flows
This research project will systematically evaluate data-driven mode decomposition techniques (POD, Spectral POD (SPOD), finite impulse response SPOD (FIR-SPOD) and a machine-learning (ML) based nonlinear method) for separating and reattaching flows using existing experimental and numerical databases of a laminar boundary layer approaching an FFS. There will provide three significant results: (1) critical insight as to the effectiveness of different dimensional reduction techniques for the analysis of separated flow dynamics, (2) the development and evaluation of an ML-based nonlinear mode decomposition method for flow analysis, and (3) enhancement of understanding of key scale interactions in separated flows. We will be able to determine which method is most suitable for enhancing our understanding the dynamics of the flow, and which is most reliable for flow estimation and control purposes, as they may not be the same. These outcomes are a step towards enhancing knowledge of the physics of separated flows which are prevalent in bluff body aerodynamics (e.g., buildings), and the improvement of data-driven flow estimation, prediction and control tools which are crucial for remote sensing applications (e.g., drones in urban environments), which are of interest to an ongoing research collaboration between Institut Pprime, Université Paris-Saclay and University of Calgary.
Voir la description complète du projetRobert Martinuzzi
Université de Poitiers
Physics
Education
University of Calgary
Globalink Research Award